We proposed a method that combines the deep learning model U-Net with a dendritic neuron model (DNM) and demonstrated its effectiveness for more robust extraction of vascular wall boundaries through analysis using ultrasound simulations and in vivo carotid artery data. Dendritic U-Net (DU-Net) is a model that replaces the output layer of U-Net with a DNM. Its accuracy was compared to U-Net using cross-validation on a dataset generated by ultrasound simulation. The applicability of the model trained on the simulation dataset to clinical data was also verified using an in vivo dataset of a healthy human carotid artery. The noise robustness of the results was evaluated using the mean absolute difference (MAD) within the scanline and frame of the extracted vessel wall boundaries from the segmentation map. Cross-validation using the simulation data showed that DU-Net achieved higher accuracy than U-Net in terms of precision (93.64
This study aimed to determine whether the addition of contrast-enhanced ultrasonography (CEUS) to B-mode ultrasonography improves diagnostic accuracy in differentiating benign from malignant breast lesions. A comprehensive literature search was performed using PubMed (MEDLINE), Ichushi, and the Cochrane Library up to July 2025. Eligible studies assessed the diagnostic performance of CEUS for suspected breast cancer, with histopathology as the reference standard. Pooled sensitivity, specificity, diagnostic odds ratio (DOR), and positive and negative likelihood ratios were calculated. Eleven studies comprising 1987 patients with 2051 lesions were included. The pooled sensitivity and specificity of B-mode ultrasonography alone were 86.1
This review aims to evaluate the clinical utility of contrast-enhanced ultrasonography (CEUS) in the differential diagnosis and management of solid and cystic pancreatic lesions. We reviewed the characteristic enhancement patterns of various solid and cystic pancreatic diseases, including pancreatic ductal adenocarcinoma (PDAC), mass-forming pancreatitis (MFP), focal autoimmune pancreatitis (AIP), pancreatic neuroendocrine neoplasms (PNENs), solid pseudopapillary neoplasms (SPNs), metastatic tumors, intraductal papillary mucinous neoplasms (IPMNs), serous cystic neoplasms (SCNs), and mucinous cystic neoplasms (MCNs). We also analyzed the role of quantitative perfusion analysis and summarized diagnostic performance based on multicenter studies and meta-analyses. In PDAC, the “irregular rolling sign” and early washout under a hypovascular pattern are critical diagnostic markers reflecting histological desmoplasia. Conversely, MFP typically shows diffuse isovascularity. CEUS facilitates the differentiation of malignant mural nodules from mucus plugs in IPMN with high accuracy. Quantitative perfusion analysis (QPA) using time-intensity curves (TIC) provides objective metrics like peak intensity (PI) and area under the curve (AUC), enhancing diagnostic objectivity. Large-scale studies, such as the Pancreatic Multicenter Ultrasound Study (PAMUS), and several meta-analyses have demonstrated that CEUS achieves a pooled sensitivity of 86–90
To evaluate whether the starry sky liver (SSL) sign is more strongly associated with blood ketone levels than with clinically defined acute gastroenteritis, and to assess the clinical utility of a quantitative ultrasound index, i.e., the SSL scale, in pediatric patients presenting with vomiting. We retrospectively analyzed 122 pediatric patients presenting with vomiting who underwent liver ultrasonography and blood testing, including blood ketone levels. The SSL scale was calculated using grayscale histogram analysis of B-mode ultrasound images. Associations between the SSL scale and clinical variables were evaluated using correlation and multivariate regression analyses. Logistic regression and receiver operating characteristic (ROC) analyses were performed to assess the ability of the SSL scale to detect elevated ketone levels (≥ 2.0 mmol/L). The SSL scale showed a significant negative correlation with log-transformed ketone levels (r = − 0.47, P < 0.001) and remained independently associated in multivariate analysis (β = − 4.87, P = 0.003). Acute gastroenteritis was not independently associated with the SSL scale. The SSL scale demonstrated good discriminative performance for detecting elevated ketone levels, with an area under the curve of 0.78 (95
This review examines artificial intelligence (AI) applications in ultrasound imaging for carpal tunnel syndrome diagnosis. Deep learning models have achieved Dice coefficients exceeding 0.85 for median nerve segmentation and diagnostic accuracy with area under the curve values up to 0.926, often matching specialist performance. However, methodological limitations exist: only three of 13 included studies (23
To determine whether prone-position transabdominal ultrasonography improves pancreatic tail visualization compared with conventional supine scanning, as assessed using computed tomography–transabdominal ultrasonography (CT–TUS) fusion imaging. This prospective observational study included 50 patients (34 males; median age, 64.5 [interquartile range, 56.5–72] years) undergoing evaluation for pancreatobiliary diseases using TUS and contrast-enhanced CT. Using Smart Fusion technology, CT and real-time TUS images were synchronized, and the pancreatic tail visualization rate was compared between the images obtained in supine and prone positions. Interobserver agreement for the measurable tail length in the prone position was assessed using the intraclass correlation coefficient (ICC: 3,1). Factors associated with pancreatic tail visualization, including sex, age, height, waist circumference, pancreatic tail length and abnormalities, and the renal–pancreatic tail distance, were analyzed using univariate and multivariate regression analyses. The pancreatic tail visualization rate, i.e., proportion of the total tail length visible on TUS compared with that on CT scan, was 37.4
Patients with recurrent or new thyroid nodules after first ablation occasionally require repeat thermal ablation (re-ablation). However, there has been no research focused on perithyroidal adhesion and complications in re-ablation of thyroid nodules. This study aimed to investigate the incidence of perithyroidal adhesion and complication in re-ablation of thyroid nodules. This retrospective self-controlled study enrolled a total of 71 patients who underwent re-ablation of thyroid nodules located in the same lobes as the first ablation. Perithyroidal adhesion at re-ablation was evaluated with hydrodissection in ultrasound images during surgery. Perithyroidal adhesion and complications at re-ablation were compared to those at first ablation. Definite adhesion was not encountered at either re-ablation or first ablation. Partial adhesion occurred in two cases (2.7
To compare short-term pain trajectories after ultrasound (US)-guided cervical selective nerve root block (SNRB) using a steroid-containing injectate versus a no-steroid injectate in patients with cervical radicular pain. This single-center prospective comparative study used time-period–based allocation of injectate. Consecutive patients undergoing a first-time US-guided cervical SNRB between 2021 and 2024 were reviewed. The steroid group received 1
To investigate the additional effects of combining electromyography (EMG) and ultrasonography (US) biofeedback with internal rotation exercise in patients with painful and restricted hand-behind-back (HBB) movement, commonly limited by subcoracoid impingement and posterior shoulder tightness. Forty-eight patients with shoulder periarthritis or subacromial pain syndrome or rotator cuff tear were randomly assigned to a control group (CG; internal rotation exercise only), an EMG biofeedback group (EMG-G), or a combined EMG and US biofeedback group (EMG + US-G). Exercises were performed daily for 4 weeks (10 repetitions × 3 sets) under weekly physical therapy supervision. The primary outcome measure was the C7–thumb distance. Secondary outcome measures included pain; Disability of the Arm, Shoulder, and Hand (DASH) score; internal rotation strength; and shoulder range of motion. Analyses were performed using an intention-to-treat approach with multiple imputation, applying mixed-effects models. A significant group-by-time interaction was observed for the C7–thumb distance (p = 0.007), with the EMG + US-G showing greater improvement compared with the CG. Pain decreased over time, with a significant group-by-time interaction (p = 0.049); however, no significant between-group differences were detected in post hoc analyses. No significant interactions were found for DASH score, internal rotation strength, or passive range of motion (p > 0.05). The addition of EMG and US biofeedback to internal rotation exercise was associated with improvement in HBB mobility in patients with limited HBB movement. The effects on pain were modest and should be interpreted cautiously, although a potential benefit cannot be excluded.
PURPOSE:Thyroid nodules are highly prevalent in the general population, yet large-scale ultrasound screening is limited by its reliance on experienced sonographers. Autonomous robotic ultrasound systems have the potential to enable standardized thyroid screening with reduced operator dependency. METHODS:In this pilot prospective study, a fully autonomous robotic ultrasound system was evaluated for thyroid nodule screening in a clinical setting. Twenty-one consecutive participants underwent both autonomous robotic ultrasound and standard operator-performed ultrasound examinations in a randomized order. The primary outcome was participant-level thyroid nodule detection rate compared with operator-performed ultrasound. Secondary outcomes included feasibility, thyroid gland coverage, and participant-reported comfort. RESULTS:Autonomous robotic ultrasound was successfully completed without human intervention in 20 of 21 participants (95.2%). Thyroid nodules were identified on operator-performed ultrasound in nine participants, of whom eight were correctly detected by the autonomous system, corresponding to a participant-level detection rate of 88.9%. Complete coverage of the left lobe, right lobe, and isthmus was achieved in 100%, 95%, and 80% of participants, respectively. Participant-reported comfort was neutral to positive, with a median Likert score of 3 (interquartile range, 3-4). CONCLUSION:This pilot study demonstrated the feasibility of fully autonomous robotic ultrasound for standardized thyroid nodule screening. The system achieved high completion rates and acceptable screening performance in a controlled clinical setting, supporting further validation in larger and more diverse populations.
Accurate evaluation of tumor viability during immunotherapy for unresectable hepatocellular carcinoma (HCC) remains challenging using contrast-enhanced computed tomography or magnetic resonance imaging (CECT/CEMRI), particularly after treatment-induced vascular changes. This exploratory study investigated whether contrast-enhanced ultrasound (CEUS) detects residual intratumoral perfusion differently from CECT/CEMRI and whether CEUS perfusion patterns are associated with depth of response (DpR) in selected patients receiving immunotherapy-based combination therapy. Twenty-five patients with unresectable HCC treated with atezolizumab plus bevacizumab, durvalumab plus tremelimumab, or nivolumab plus ipilimumab who underwent both CEUS and CECT/CEMRI within a 3-month interval were retrospectively analyzed. CEUS was performed at a median of 33 days after treatment initiation. Arterial-phase enhancement patterns were classified as hyper-, iso-, or hypoenhancement relative to surrounding liver parenchyma. Tumor response was evaluated using RECIST version 1.1 and modified RECIST. DpR was defined as the maximum percentage reduction in tumor size from baseline. CEUS demonstrated hyperenhancement in 14 lesions compared with six lesions on CECT/CEMRI. Enhancement patterns were discordant in 13 cases, with CEUS demonstrating hyperenhancement in 10 lesions classified as iso- or hypoenhancement on CECT/CEMRI (p = 0.049). Tumor size reduction did not differ significantly among enhancement categories on CECT/CEMRI (p = 0.325), whereas CEUS enhancement patterns were significantly associated with DpR (p = 0.008). In selected patients with unresectable HCC undergoing immunotherapy-based therapy, CEUS detected residual intratumoral perfusion more frequently than CECT/CEMRI and was significantly associated with DpR. CEUS may provide complementary information for exploratory assessment of tumor viability in this setting.
To determine whether a dynamic ultrasound approach, termed anatomical scanning and interpretation, improves diagnostic performance in differentiating benign from malignant breast lesions compared with conventional static-image BI-RADS assessment. This prospective study evaluated 102 breast lesions (33 masses, 41 non-masses, and 28 distortions) classified as BI-RADS 3, 4a, or 4b. Fourteen ultrasound technologists underwent dedicated training in dynamic anatomical scanning and interpretation. Each lesion was evaluated using real-time video clips and independently categorized as requiring follow-up or biopsy. A subsequent physician–technologist consensus was performed. With histopathology as the reference standard, performance was assessed using ROC analysis, with AUC comparisons performed using the DeLong test. Among 102 lesions (58 benign, 44 malignant), individual technologist interpretation with dynamic imaging significantly improved diagnostic performance (AUC = 0.748, p = 0.0191) over static BI-RADS assessment (AUC = 0.619). Physician–technologist consensus showed further significant improvement (AUC = 0.865, p < 0.0001). In subgroup analysis, dynamic interpretation outperformed static assessment for masses (AUC = 0.849, p = 0.0036), but not for non-mass lesions (AUC = 0.794) or distortions (AUC = 0.618). In contrast, physician–technologist consensus interpretation demonstrated superior performance compared with static assessment across all lesion types: masses (AUC = 0.879, p < 0.0001), non-mass lesions (AUC = 0.910, p = 0.0005), and distortions (AUC = 0.838, p = 0.0432). Dynamic anatomical scanning and interpretation significantly improve diagnostic performance over conventional static BI-RADS assessment. For non-mass lesions and architectural distortions, anatomical interpretation based on consensus between the physician and the technologist is considered essential to achieve optimal diagnostic accuracy.
PURPOSE:Pulse inversion (PI) harmonic imaging has been commonly used because of its contrast and spatial resolution superior to conventional fundamental imaging. However, it requires two transmissions with opposite phases, which reduces the imaging frame rate by half. This study aimed to solve this problem METHOD: An ultrasonic array probe consisting of piezoelectric elements with two polarization-inverted layers, which requires only one transmission to extract the second harmonic components, was developed. Its characteristics, such as electric impedance and its phase, were predicted by solving piezoelectric as well as wave equations, and then compared with measurement. The acoustic pressures produced from the developed probe, when excited at the lower and higher resonant frequencies, were measured using a hydrophone. Finally, imaging of wires in water and phantoms containing wires and a cyst model were performed using the developed probe RESULT: The spatial resolution, especially in the axial direction, was significantly improved with the developed probe consisting of the two piezoelectric layer elements in comparison with the conventional second harmonic imaging (HI) using a band pass filter. It was almost comparable to that of conventional PI imaging despite only one transmission being required CONCLUSION: A probe consisting of two piezoelectric layer elements enabling one-transmission HI was developed. The spatial resolution obtained using the developed probe was slightly worse but almost comparable to that of conventional PI imaging even with only one transmission.
The increasing prevalence of steatotic liver disease (SLD) underscores the need for reliable, noninvasive tools to assess hepatic steatosis. Existing methods, such as the controlled attenuation parameter (CAP) under ultrasound, have limitations in accuracy. This study aimed to examine whether combining attenuation imaging (AI) and hepatorenal index (HRI) measurements improves discrimination of steatosis in participants with chronic liver disease (CLD). This multicenter prospective cohort study enrolled 121 participants with CLD from two centers during 2022–2023. All participants underwent liver biopsy and ultrasound examinations, including AI and HRI measurements using a Philips EPIQ system, and CAP assessment. An AI-HRI composite score was developed using logistic regression and evaluated against individual variables through receive operating characteristic analysis, category-free net reclassification improvement (cf-NRI), and integrated discrimination improvement (IDI). Internal validation was performed using bootstrap sampling. The area under the curve values for discrimination for the AI-HRI composite score, AI, HRI, and CAP were 0.91, 0.87, 0.86, and 0.80, respectively. At a cutoff of -0.083, the sensitivity and specificity were 84.1
